The work outlines our vision and implementation details for converting expert knowledge in natural language into an ontology graph. The approach is used for step-by-step learning from experts by reproducible game tree solver in learning expert meaning processing. Having an ontology graph gives the advantage of easier knowledge integration and reasoning using tools like the SPARQL query language. Our approach is based on a combination of natural language processing and transformers, i.e., spaCy and BERT; and semantic web, i.e., resource description framework, web ontology language and semantic web rule language. The proposed solution consists of the following steps: (1) preprocessing text for named entity recognition, (2) contextual understanding and relation extraction, (3) mapping extracted triples into new layers of knowledge in web ontology language/resource description framework, and (4) defining and extracting rules and mapping them into semantic web rule language.
Grigoryan et al. (Mon,) studied this question.